Towards accuracy recognition and content estimation of typical pesticides in groundwater via electronic nose

Accurate and rapid prediction of pesticides in groundwater during groundwater remediation is important to optimize remediation processes and increase remediation efficiency. In this paper, an electronic nose was employed to recognize the pesticides in groundwater. Four typical pesticide-polluted and unpolluted groundwater samples were prepared and their odor information was collected by e-nose. The odor signals collected by the e-nose combined with machine learning algorithms were used to predict pesticides in groundwater. Firstly, based on different feature extraction methods, random forest was used to distinguish whether groundwater was polluted by pesticides. The recognition rate of both the training set and the test set were 100%. Secondly, the random forest was applied to determine the type of pesticides in the polluted samples, with recognition rates of 100% and 99.29% for the training and test sets, respectively. Finally, the concentrations of the four typical pesticides were predicted by the support vector regression models. The correlation coefficient R2 of the training set model was 0.99, and that of the test set ranged from 0.94 to 0.99. From the case studied, the e-nose technique can be applied in the recognition of pesticide in groundwater, which will contribute to the targeted remediation of groundwater.